The hidden value of air transportation infrastructure

Technological Forecasting & Social Change 74 (2007) 18–35

The hidden value of air transportation infrastructureB

Bruno Miller a,*, John-Paul Clarke b,1

a 4313 42nd Ave. S., Minneapolis, MN 55406, USAb School of Aerospace Engineering, Georgia Institute of Technology, 270 Ferst Drive NW, Atlanta,

GA 30332-0150, USA

Received 31 July 2003; accepted 31 March 2004

Abstract

Air transportation is a key strategic asset in that it provides access to markets and thereby enables the

economic development of nations. Thus, in order to maintain their competitiveness in a global economy,

countries must invest in air transportation infrastructure to ensure their ability to meet current and future

demand for aviation services. The objective of this paper is to develop and illustrate a methodology for

evaluating the strategic value of air transportation infrastructure, in particular the benefits associated with

the ability to react quickly to changes in the market. The hypothesis is that by recognizing and taking

advantage of this strategic value, it may be possible to design better policies for aviation infrastructure

delivery.

The methodology developed here uses system dynamics to model different strategies for infrastructure

delivery. These strategies are defined by three variables: the amount of capacity increase, the time to

deliver the capacity and the congestion threshold that triggers the need for capacity delivery. Monte

Carlo simulation is used to take into account multiple sources of uncertainty. The model shows that a

strategy of capacity delivery based on small increments and short response times can yield more benefits

than strategies that consider large capacity increases and long response times. Furthermore, in the specific

airport example considered here, it was found that a congestion threshold of 75% should be the trigger

for capacity enlargements if strategies based on small capacity increments and 1 or 5 years to increase

0040-1625/$ -

doi:10.1016/j.t

B Presented a

2003. Selected

* Correspond

E-mail add1 Tel.: +1 40

see front matter D 2006 Published by Elsevier Inc.

echfore.2004.03.011

t the 7th International Conference on Technology Policy and Innovation in Monterrey, Mexico, June 10th–13th,

for publication in a special volume of the journal Technological Forecasting and Social Change.

ing author. Tel.: +1 617 291 6352.

resses: [email protected] (B. Miller)8 [email protected] (J.-P. Clarke).

4 385 7206.

http://dx.doi.org/10.1016/j.techfore.2004.03.011

B. Miller, J.-P. Clarke / Technological Forecasting & Social Change 74 (2007) 18–35 19

capacity are considered. The lesson for decision-makers is that congestion delays must be addressed with

foresight.

D 2006 Published by Elsevier Inc.

Keywords: Investments in air transportation infrastructure; Strategic value of infrastructure; Air transportation

1. Introduction

Air transportation is a key strategic asset in that it provides access to markets and thereby enables

the economic development of nations and regions [1]. Thus, in order to maintain their

competitiveness in a global economy, countries must invest in air transportation infrastructure to

ensure their ability to meet current and future demand for aviation services.

Investments in air transportation infrastructure are difficult to evaluate for a variety of reasons, but

two are especially worth noting. First, investments in aviation infrastructure must be considered as part

of a wider national strategy that involves many other sectors [2]. Second, the size of these investments

is typically large and the time for implementation is also usually long. Thus, the benefits of the

investment may not be realized for many years [3]. Given these difficulties, investments in air

transportation infrastructure are often not implemented or are implemented too late relative to the

needs of society.

The methodology presented in this paper is based on the assumption that it is useful to consider

investments in aviation infrastructure not as an end in themselves, but rather as a way to provide

options to accommodate future growth scenarios. In general, current thinking considers the benefits of

infrastructure in terms of the function it was built to perform; however, there are at least three other

types of benefits that infrastructures can provide: a) the ability to respond quickly to changes in the

market, b) the ability to explore markets, and c) the ability to influence markets. By considering the

strategic value of these three benefits, decision-makers can better understand the overall benefits of

the infrastructure and justify its delivery.

2. Objective

The objective of this paper is to develop and illustrate a methodology for evaluating the strategic

value of air transportation infrastructure, i.e. the benefits associated with the ability to react quickly to

changes in the market, to explore new markets and to stimulate existing markets. The hypothesis is

that by recognizing and taking advantage of this strategic value, it may be possible to design better

policies for aviation infrastructure delivery.

The simulation developed in this study focuses primarily on capturing the benefits due to the

ability to react quickly to changes in the market. The simulation can be enhanced to also

incorporate the benefits from exploring and stimulating markets. These enhancements have not

been elaborated here mostly to simplify the calculations and the presentation of the results;

however, at the end of the paper, we indicate how the model could be expanded to incorporate

these other benefits.

B. Miller, J.-P. Clarke / Technological Forecasting & Social Change 74 (2007) 18–3520

The next section explains the methodology employed to analyze the strategic value of air

transportation infrastructure. Then, numerical results and sensitivity analysis for the specific example

considered here are presented. The paper concludes with a summary of the main points and a discussion

of the implications of the results for policy-making.

3. Methodology to calculate the strategic value of aviation infrastructure

The methodology presented here is based on a combination of system dynamics and Monte Carlo

simulation. The first step is to model the infrastructure delivery problem using system dynamics tools.

The main advantage of this approach is that it captures the feedback loops and delays that occur in

complex systems, such as air transportation. The next step is to determine different strategies for

infrastructure delivery and implement them in the system dynamics model. Finally, Monte Carlo

simulation is used to take into account multiple sources of uncertainty.

3.1. Modeling the infrastructure delivery problem with system dynamics

3.1.1. Overview of the model

The example considered in this study is a hypothetical yet common situation in many airports around

the world. Here, it is assumed that an element of the air transportation infrastructure (e.g. a runway or a

passenger building) is reaching saturation, and, thus, depending on the level of traffic, may lead to

congestion (see Fig. 1).

In this particular example, Runway capacity2 is the limiting factor that leads to congestion. As

demand for air travel (Aircraft per hour) increases, the total number of aircraft requesting service on this

runway (Total aircraft) also increases. If Runway capacity is held constant, the increase in demand

slowly leads to Congestion, which raises the direct operating costs of airlines (Airline congestion cost).

The higher operating costs are passed on to the passenger in terms of higher airfares (Airfare impact)

and this leads to less demand for travel (Congestion cost loop). In addition, congestion decreases the

level of service by lengthening passenger travel time (Level of service impact) which also results in

less demand for aviation services (Passenger comfort loop). Once Congestion reaches a certain limit

(Congestion threshold), a certain amount of capacity (Capacity increase) is delivered after a certain

period of time (Years to increase capacity). These three variables (Congestion threshold, Capacity

increase, and Years to increase capacity) are specified by the user and they characterize the different

infrastructure delivery strategies. Once capacity is added to the runway, Congestion decreases, thus,

stimulating demand by reducing the Airfare impact and Level of service impact. The model assumes

that congestion occurs only at a given number of Peak hours per year.

There are two main outputs from this model. The first are the benefits accrued to the airport

operator in terms of Airport revenues. Here, it is assumed that they consist mainly of Landing fees

paid by the airlines and passenger facility charges (PFC) paid by each traveler. The second output is

the cost of infrastructure construction (Delivery cost). Together, these two outputs determine the cash

flows of the airport operator and form the basis for calculating the net present value (NPV) for the

different strategies.

2 Model variables are written in Italics throughout the whole paper.

Fig. 1. System dynamics model of the hypothetical situation considered in this study.


3.1.2. Main relationships of the model

The main relationships used in the model are explained below:

a) Congestion

Congestion in this model is defined as the waiting time, Wq, for each user (aircraft) that wants to land

on the runway. The waiting time is obtained by modeling this situation as a M/G/1 queuing system (i.e. a

system with Poisson demand, any type of service time, one server and infinite capacity) [4]:

Wq ¼kd 1

l

� �2þ r2

t

� �

2d 1� qð Þ ð1Þ

where k is the average demand rate over a specified period of time determined by the Poisson

distribution, l is the capacity of the system over the same period of time, rt is the standard deviation of

service times and q is the butilization ratioQ. The metric for congestion used in this study is bhours perpeak hourQ, i.e. hours of waiting time per peak hour of traffic.

The utilization ratio q can be defined as the average demand rate over system capacity [5]:

q ¼ kl

ð2Þ

In the model, k is given by Total aircraft and l is Runway capacity. A key characteristic of this

queuing system is that as k approaches l, i.e. as demand approaches capacity, the numerator of Eq. (1)

50%

55%

60%

65%

70%

75%

80%

85%

90%

95%

98%

Utilization ratio

Co

ng

esti

on

[hr/

hr]

0.00

0.10

0.20

0.30

0.40

0.50

0.60

Fig. 2. Waiting time (in terms of bhours of waitingQ per hour) of an M/G/1 system as a function of utilization ratio for the

parameters (k, l and r) chosen in this model.


tends to zero and the waiting time grows non-linearly [5]. For our example, this means that when aircraft

demand per hour approaches runway capacity per hour, and, hence the utilization ratio approaches 1,

Congestion becomes critical (see Fig. 2):

b) Annual demand growth rate

Demand for air travel is very volatile and cyclical [6]. In order to capture this behavior, Annual

demand was modeled as a mean reverting stochastic process of the form shown in Eq. (3):

dx ¼ gd X � xð Þdt þ rddz ð3Þ

where dx is the change in demand x over a time interval dt, g is the speed of reversion, i.e. a metric that

represents how fast the process returns to its long-term trend, X is the level to which x tends to revert, r is

the variance parameter and dz represents a Wiener process, i.e. a normally distributed random process (for

more details, please consult [7]). This equation was calibrated with historical data for air travel demand in

the Unites States between 1979 and 2001 contained in the Form 41 database [8] and using the method

Annual growth rate0.2

0.1

0

-0.1

-0.20 5 10 15 20 25 30 35 40 45 50

Time (Year)

Annual growth rate : sample run Dmnl

Annual growth rate : sample run 2 Dmnl

Fig. 3. Sample paths for air travel demand growth rate as modeled in this study.


outlined in Ref. [7]. The long-term annual demand growth rate found was on the order of 3.8% per year.

Two sample paths of this mean reverting process are illustrated in Fig. 3:

c) Airfare impact and Level of service impact

The Airfare impact and the Level of service impact on air travel demand were determined using the

concepts of price and time elasticities, respectively [9]. The price elasticity of demand is defined as the

percentage change in total demand that occurs with a 1% change in average airfare. Similarly, time

elasticity of demand is the percentage change in total travel demand that occurs with a 1% change in

travel time. In this study, the Airfare impact is the change in demand from a percentage change in

average travel cost times price elasticity (see Eq. (4)):

Airfare impact ¼ epricedDCostTravel ð4Þ

where eprice is the price elasticity of demand and DCostTravel is the percentage increase in travel cost due

to the congestion costs that the airlines are able to pass on to the consumers. Historically, price elasticity

of demand has been estimated at approximately �1.6 for leisure passengers and on the order of �0.8 forbusiness passengers [9]. DCostTravel is given by Eq. (5):

DCostTravel ¼CostCongestion=passenger

AirfareAveragedCostTransfer ð5Þ

where CostCongestion/passenger is the average airline operating cost per hour divided by an average

number of passengers per flight, AirfareAverage is the average air fare charged per passenger, and

CostTransfer is the percentage of CostCongestion that the airline is able to pass on to the travelers.

The Level of service impact is the change in demand from a percentage change in average travel time

times time elasticity (see Eq. (6)):

Level of service impact ¼ eTimedDTimeTravel ð6Þwhere etime is the time elasticity of demand and DTimeTravel is the percentage change in travel time due to

congestion. Time elasticity of demand is nominally considered to be around �0.8 for leisure passengers

and approximately �1.6 for business passengers [9].

d) Capacity delivery costs

In this model, airport capacity is given in terms of aircraft per hour. Capacity delivery costs are

obtained by multiplying Capacity increase times a unit capacity delivery cost in US$/aircraft per hour.

Here, the unit capacity delivery cost is assumed to be US$5 million/aircraft per hour. For example,

this means that an increase of 20 aircraft per hour (i.e. 50% of the assumed runway capacity of the

airport in the model) requires an investment of US$100 million. Runway construction costs vary

substantially from airport to airport, ranging anywhere from US$60 million to US$300 million, and

even more in some particular cases [10]. Thus, the unit capacity delivery cost of US$5 million/aircraft

per hour is just meant to be an approximation to costs in the real world.

e) Congestion threshold, Capacity increase, Years to increase capacity

These constants are the main control parameters to specify the different infrastructure development

strategies. Congestion threshold is the maximum level of congestion allowed before the airport

operator decides to add more capacity. Capacity increase is the amount of capacity to be added once

the decision has been taken to deliver more capacity. Finally, Years to increase capacity determine the

delay in supplying the desired capacity once the decision has been made to deliver it.


3.1.3. Model calibration

The numbers used to calibrate the model are meant to illustrate a realistic situation but they do not

represent an actual airport. The airport in this study is assumed to be a one runway facility that

serves primarily narrow-body aircraft. The current runway capacity was set at 40 aircraft per hour.

Landing fees were estimated at $200 per aircraft based on data in Ref. [5] and the typical weight of

narrow-body aircraft. It is further assumed that congestion occurs only at peak hours and there are

1000 peak hours in a year. The simulation time period is in years and each run covers 50 years.

3.1.4. Infrastructure delivery strategies

For the purposes of this study, each infrastructure delivery strategy is defined by three parameters:

1) Congestion threshold

This constant determines the maximum level of congestion allowed (btriggerQ) before the airport

operator decides to put more capacity in place. A low threshold reflects a proactive strategy by which the

operator intends to have enough capacity to meet demand and, thus, be able to capture more revenues.

The disadvantage of such approach is that if demand does not grow as expected, the new capacity may

remain unused. A high threshold represents a reactive strategy. Here, an airport operator would wait until

it is obvious that the current levels of demand require more capacity. The disadvantages of this strategy

are the negative effects of congestion on travel demand explained above.

Each congestion threshold corresponds to a given utilization ratio, q, as determined by Eq. (1) and

shown in Fig. 2. Five congestion thresholds were considered and are presented in Table 1.

2) Capacity increase

Another strategic decision by the airport operator is how much capacity to add. In order to simplify

the analysis, only three cases were considered: 10%, 25% and 50% increase of existing capacity. The

first alternative (10%) represents a very incremental approach, in which little additions to capacity are

made at any given time. For example, modifications to existing approach procedures and/or better

sequencing of arriving aircraft enabled by automation support tools may lead to these small increments.

The second alternative (25%) may include the installation or upgrade of Instrument Landing Systems

(ILS) and/or expansion of taxi-ways and holding areas, for example. The third case (50%) represents a

larger investment like a completely new taxi-way or runway.

The advantage of small capacity increments is that demand may be matched more closely to capacity,

thus, avoiding the delivery of unnecessary infrastructure. In addition, capital expenditures are smaller

and spread over time, which may be easier to finance and justify than large investments performed at

once. Even though it may not be entirely realistic to have a purely incremental strategy throughout the

life of an airport, as it is not possible to add incremental capacity by building little pieces of runway at a

time, for example, this strategy illustrates a limiting case. The medium and large capacity increments

Table 1

Relationship between congestion threshold and utilization ratio for the example analyzed in this study

Congestion threshold (hr/peak hr) Utilization ratio, q (%)

0.019 60

0.029 70

0.038 75

0.113 90

0.239 95

Table 2

Hypothetical relationship between Years to increase capacity and level of preparedness for the process of building a runway

Years to increase capacity Level of preparedness

A priori investments Outstanding tasks

1 – Purchase land – Install navigation aids

– Permitting process completed

– Runway is built

5 – Own land – Build runway

– Permitting process completed – Install navigation aids

10 – Purchase land – Permitting process

– Build runway

– Install navigation aids

15 – Purchase land

– Permitting process

– Build runway

– Install navigation aids


lead to more significant investments at particular points in time. These are more representative of what is

seen in practice, as airport capacity enhancements projects tend to be large.

3) Years to increase capacity

The time that it takes to increase capacity reflects how fast the airport operator can react to changes in

the market place. Four cases are considered in this study: 1, 5, 10 and 15 years. It is assumed that the

time to increase capacity depends on the blevel of preparednessQ of the operator to respond to changes. Ashort time to increase capacity means that the airport operator has invested resources a priori to prepare

its facilities for expansion. If we consider the example of building a new runway, the different Years to

increase capacity may correspond to the level of preparedness shown in Table 2.

The implication with these assumptions is that in order to be able to respond faster, the operator of the

facility must have invested some resources beforehand. Two important considerations arise from this

observation. First, the level of a priori investment depends on the benefits that reacting quickly brings.

Consequently, if the purpose of the a priori investment is to maximize the economic returns of the whole

project, the maximum amount that somebody should be willing to invest a priori for the ability to react

quickly is the difference between the benefits derived from reacting quickly compared to the benefits

under the bbusiness-as-usualQ scenario.The second consideration is that the a priori investment may lower the cost of reacting quickly in the

future. For example, buying land for another runway may be less expensive before traffic levels require

the addition of this extra runway. If the airport waits until its demand requires another runway, the

current owner of the land intended for the extra runway is likely to increase the price of the land because

Table 3

Differences in capacity delivery cost for the Moderate cost difference scheme

Years to increase capacity Reduction in capacity delivery cost/unit of capacity (%)

1 �505 �2510 0

15 +25

Capacity increase (% of current capacity)

1

5

10

15

60% 70% 75% 90%

10% 25%

50%

Years to increasecapacity(Years)

95%

Congestion threshold(utilization ratio)

Fig. 4. Matrix of the delivery strategies analyzed in the study.


this situation represents a seller’s market. Consequently, the premium associated with the a priori

investment to obtain the ability to react quickly may be compensated with lower capacity delivery costs.

In order to illustrate this point, two different construction cost schemes were considered. The first one,

the so-called bZero cost differenceQ, assumes no difference in capacity delivery costs. The second one,

the so-called bModerate cost differenceQ, assumes the following reductions in capacity delivery costs per

unit of capacity (see Table 3).

Notice that there is no difference in delivery cost for 10 years to increase capacity because this is the

bbusiness-as-usualQ scenario. In addition, it is assumed that having a longer time to deliver capacity (15

years) increases the delivery costs.

3.1.5. Summary of delivery strategies

The study considered all possible combinations of the above parameters (Congestion threshold,

Capacity increase and Years to increase capacity) for a total of 60 delivery strategies (see Fig. 4). Even

though it may lead to some unlikely scenarios, such as adding 50% of capacity every year, the test

matrix does span a representative sample of realistic strategies with some limiting cases. Typical

strategies seen in the real world arguably fall between 25% and 50% Capacity increase with 10 to 15

Years to increase capacity, with a Congestion threshold of 90% to 95%.

3.1.6. Monte Carlo simulation

Monte Carlo simulation is used to incorporate multiple sources of uncertainty into the calculations.

Here, the main uncertainties considered were related to Annual demand growth rate, Air fare impact and

Level of service impact. Table 4 shows the variables selected for the Monte Carlo simulation along with

their assumed probability distributions and limiting values.

Table 4

Variables considered for the Monte Carlo simulation and their assumed probability distributions

Variable Units Probability distribution Min. value Max. value

Annual demand growth rate seed N/A Uniform 1 10,000

Average travel time Hours Uniform 2 4

Time elasticity N/A Uniform �1.6 �0.8Price elasticity N/A Uniform �1.6 �0.8Costtransfer % Uniform 60% 90%


As explained above, Annual demand growth rate is modeled stochastically. In order to get different

random distributions of this variable, the seed parameter is changed between the limits shown in Table

4. Average time refers to the assumed average trip time to the hypothetical airport, which is the basis to

calculate the time sensitivity of demand and, hence, the Level of service impact. The time and price

elasticities determine the impact of congestion on travel demand because of increased travel time and

higher airfares. The elasticities are assumed to be between �1.6 and �0.8 [9]. Costtransfer reflects the

proportion of the increased operating costs due to congestion that the airlines are able to pass on to the

passengers. It is assumed that this value oscillates between 60% and 90%.

A Monte Carlo simulation was run for each of the delivery strategies shown in Fig. 4. Each

simulation consisted of 7500 runs. The output of each simulation is a distribution of Airport revenues

and Delivery costs, which are then used to calculate the net present value of each delivery strategy.

3.1.7. Evaluation of the delivery strategies

The Monte Carlo simulation generates a distribution of Airport revenues and Delivery costs for

each delivery strategy considered. For each run in the simulation, the present value of Airport

revenues and Delivery costs is calculated, thus, a distribution of present values of revenues and

present values of costs is obtained. The expected net present value for each delivery strategy i is the

difference between the mean of the present value of revenues minus the mean of the present value of

costs (see Eq. (7)):

Expected NPVi ¼ E Present Value ðRevenuesÞi� �

� E Present ValueðDelivery CostsÞi� �

ð7Þ

The standard deviation of the net present value for each strategy i is given by Eq. (8):

ri ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffir2Revenues;i þ r2

Del:Costs;i þ 2dCovRevenues;i;Del:Costs;i

qð8Þ

where rRevenues,i and rDel.Costs,i are the standard deviations of the present values of Airport revenues and

Delivery costs for strategy i, respectively, and CovRevenues,i; Del.Costs,i is the covariance between the

present value of Airport revenues and Delivery costs for strategy i.

The calculation of the present values requires the selection of a discount rate. For the purposes of this

study, a discount rate of 10% was chosen. There is always some uncertainty about what the discount rate

should be, therefore, sensitivity analyses were also performed.

In order to determine the strategic value of being able to react quickly, the expected NPV for each

delivery strategy was compared to a baseline case that exemplifies a bbusiness-as-usualQ situation. Thebaseline case was assumed to be the strategy were 50% of the current capacity is delivered every 10

years after a congestion threshold of 90% has been reached. Consequently, the expected value of quick

reaction for strategy i is given by Eq. (9):

Expected Value of quick reactioni ¼ Expected NPVi � Expected NPVBaseline ð9Þ

Remember that the expected value of quick reaction is also the maximum amount of a priori

investment that an airport operator is willing to spend to implement strategy i. The standard deviation of

the expected value for strategy i is given by Eq. (10):

rExpected Value of quick reaction ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffir2i þ r2

Baseline

qð10Þ


where ri and rBaseline are the standard deviations of the expected NPV of strategy i and the expected

NPV of the baseline scenario, respectively.

For delivery strategies that consider 15 years to increase capacity, Eqs. (9), (10) would compute

the bexpected value of delayed reactionQ and its standard deviation, respectively, as opposed to the

value to react quickly. These cases are considered here because there may be advantages to delaying

investments.

4. Results

4.1. Zero cost difference

The results for the situations where there are no cost difference for delivering capacity quicker are

shown below. For the case of 1 year to deliver capacity, the results suggest that strategies involving small

increments (10% capacity increase) lead to higher expected values to react quickly than strategies with

larger capacity increments (see Fig. 5). For example, for a Congestion threshold of 90%, the ability to

react quickly provides benefits on the order of $14.2 millionF14 million. Thus, the airport operator

should be willing to spend up to this amount a priori to have the ability to react quickly. Large increases

in capacity (25% or 50% capacity increase) generate, on average, negative expected values for the ability

to react quickly. The intuition behind this is that larger expansions may result in over-capacity that is not

utilized until some years after it has been delivered and, thus, the investment is more difficult to recover.

The results further show a large standard deviation and a strong dependence on the Congestion

threshold. The large standard deviation can be explained by the fact that since capacity is being delivered

very quickly, the system is better able to accommodate increases in demand. This is a reinforcing loop in

which the system delivers more capacity as demand grows, thereby stimulating even more traffic. When

the time to deliver capacity is longer, the system cannot react as quickly to changes in demand and,

therefore, the variability of traffic is reduced. The dependence of the results on the Congestion threshold

Fig. 5. Results showing the expected value of quick reaction for strategies with Years to increase capacity equal to 1 as a

function of increases in capacity and utilization ratio (Congestion threshold) with the assumption that there is no difference in

the delivery costs.

Fig. 6. Results showing the expected value of quick reaction for strategies with Years to increase capacity equal to 5 as a


the delivery costs.


arises from the way that present values are determined. Higher Congestion threshold means that capital

expenditures come further in the future and, thus, they get discounted more. The extra revenues from an

earlier capacity expansion are not large enough to offset the higher present cost of the earlier investment.

Therefore, with higher Congestion threshold and 1 year to increase capacity, the expected value of the

ability to react quickly increases.

As the time to deliver capacity increases from 1 to 5 years, small capacity increments still lead to

higher expected values than larger capacity increments (see Fig. 6). Notice, however, that the standard

deviation and the dependence on the Congestion threshold are less than in the case of 1 year to increase

capacity. Because the response time to deliver capacity is longer, demand growth has is limited and,

therefore, it does not vary as much as when capacity can be delivered within a year. Furthermore, the

longer time to add capacity allows demand to grow to levels that make more use of the new capacity.

Fig. 7. Results showing the expected value of delayed reaction for strategies with Years to increase capacity equal to 15 as a


the delivery costs.


Therefore, on average, the expected value to react quickly is larger for strategies that have 5 years to

increase capacity compared to strategies with 1 year to increase capacity.

For longer times to deliver capacity (15 years), similar observations as in the previous cases can be

made. Smaller capacity increases lead to higher expected values for the ability to delay reaction than

larger capacity increases (see Fig. 7). Furthermore, the variability of the results and the dependence on

the Congestion threshold are also less. Notice that the expected value of delayed reaction is positive in

almost all cases. Because it takes so long to deliver the new capacity, demand is very likely to grow to

levels where the added capacity will be completely utilized in almost all cases, thus, recovering the

investment very quickly.

In summary, for a capacity delivery scheme in which there is no cost reductions associated with

delivering capacity quickly, the strategy that results in the highest value for the ability to react quickly is

to increase capacity in small increments (10% capacity increase) with 5 years to deliver capacity,

regardless of the congestion threshold. Lower time to increase capacity, i.e. 1 year, may lead to positive

values for the ability to react quickly only if the congestion threshold is high. For strategies with 15 years

to deliver capacity, the expected value of delayed reaction is similar to the expected value of the ability to

react quickly with 5 years to deliver capacity, independent of the congestion threshold.

4.2. Moderate cost difference

The results for the cases where there is a moderate difference in the delivery costs per unit of capacity

for strategies with lower time to increase capacity are shown below. For example, for strategies with 1

year to deliver capacity, the ability to react quickly becomes very valuable in all cases (see Fig. 8).

Similarly to what was shown in the previous sub-section, the expected value of the ability to react

quickly increases and has lower variability with smaller capacity increases.

Notice that the expected value of quick reaction depends significantly on the Congestion threshold. In

fact, there seems to be an optimal Congestion threshold of approximately 75% utilization ratio for all of

the capacity increases with 1 year to deliver capacity. This is explained by the observation that

Fig. 8. Results showing the expected value of quick reaction for strategies with Years to deliver capacity equal to 1 as a function

of increases in capacity and utilization ratio (Congestion threshold) with the assumption that there is moderate difference in the

delivery costs.

Fig. 9. Results showing the expected value of quick reaction for strategies with Years to deliver capacity equal to 5 as a function

of increases in capacity and utilization ratio (Congestion threshold) with the assumption that there is moderate difference in the

delivery costs.


Congestion thresholds lower than 75% probably lead to more capacity delivered than what is needed to

meet demand, thus resulting in expensive over-capacity. Moreover, higher Congestion thresholds imply

that the negative effects of increased congestion are larger than the savings in present delivery costs

accrued by delaying the investment. Thus, if it takes 1 year to deliver capacity, the optimal strategy for

an airport operator would be to increase capacity in small to medium increments (10% or 25%) with a

Congestion threshold of 75%.

Similar observations can be made when there are 5 years to deliver capacity (see Fig. 9). Here, the best

strategies would also be to deliver small or medium increments of capacity with a Congestion threshold

Fig. 10. Results showing the expected value of delayed reaction for strategies with Years to deliver capacity equal to 15 as a

function of increases in capacity and utilization ratio (Congestion threshold) with the assumption that there is moderate

difference in the delivery costs.


of 75%. Notice, however, that the expected values to react quickly in these cases are much lower than in

the case when there is 1 year to deliver capacity. This is probably a reflection of the assumption that the

delivery cost per unit of capacity is less for strategies with 1 year to increase capacity; however,

remember that in order to gain the ability to react within 1 year, the airport operator would also have to

make a larger a priori investment. Thus, the net benefits of both strategies may be comparable.

For strategies with 15 years to deliver capacity, the value of delayed reaction is positive for small and

medium increments in capacity (see Fig. 10). Contrary to the two cases discussed above, there is not an

optimal Congestion threshold here. The only observation that can be made is that a higher Congestion

threshold seems to yield higher values of delayed reaction. Notice that, overall, the expected value of

delayed reaction is much lower than the expected value to react quicklywith 1 or 5 years to deliver capacity.

In summary, if there are moderate cost reductions associated with faster capacity delivery, the

optimal policies would be to deliver small to medium capacity increases with a congestion threshold

of approximately 75%. The strategies with 1 year to deliver capacity generate the most expected

value of quick reaction, but the net benefits may be comparable to strategies with 5 years to increase

capacity because the a priori investment for being able to react within 1 year should be larger.

5. Sensitivity analysis

The main reason to perform sensitivity analysis in this study is to determine the effect of changes in

the discount rate on the expected values of the ability to react quickly or delay reaction. The other

principal sources of uncertainty are taken into account in the Monte Carlo simulation and require no

further testing. The sensitivity analysis measures the percentage change on the expected value to react

quickly or delay reaction due to a 1% change in the discount rate. A sensitivity on the order of 1 means

that the results are very sensitive and a sensitivity of zero denotes no impact of the discount rate on the

results. A negative sensitivity implies that the change is in the opposite direction as the change in the

discount rate.

The sensitivity analysis was performed by considering the effect of a �20% and +20% change in

the discount rate, i.e. changes due to a discount rate of 8% and of 12%, respectively (recall that the

Fig. 11. Effect of a discount rate of 8% on the value to react quickly for all strategies considering 5 years to deliver capacity.

Fig. 12. Effect of a discount rate of 12% on the value to react quickly for all strategies considering 5 years to deliver capacity.


baseline discount rate was 10%). The analysis was performed on all strategies with 5 years to deliver

capacity.

Fig. 11 shows the sensitivity to a discount rate of 8%. In general terms, the sensitivity of the

results is very high and this sensitivity increases with larger capacity increments and with decreasing

congestion thresholds. For small capacity increments, the sensitivity is on the order of �0.9 for low

congestion thresholds and 0.1 for a congestion threshold of 95%. Large capacity increments show

high sensitivity, especially the case for a congestion threshold of 60% which has a sensitivity on the

order of �35.It is not surprising that the larger capacity increments are more sensitive to the discount rate

because they imply the largest capital expenditures and, thus, their absolute change in magnitude is

larger. Lower congestion thresholds mean that there may be more investments, all else held equal, and,

therefore, the absolute change in the magnitude of the investments will also be larger.

The sensitivity to a discount rate of 12% is shown in Fig. 12. The same observations as for the case of

an 8% discount rate hold: sensitivity is very high and it increases with larger capacity increments and

lower congestion thresholds.

The sensitivity analysis clearly shows that the choice of discount rate is of critical importance. This

is an aspect that deserves further investigation; however, rather than a weakness of the system

dynamics and Monte Carlo simulation methodology presented here, the high sensitivity to the discount

rate is a reflection of the great care that must be taken when using discounted cash flow and net

present value to evaluate investments. Furthermore, this analysis shows that larger investments are

more sensitive to the discount rate than smaller investments, which may be a further argument towards

smaller-scale capacity enhancements.

6. Conclusions

The model results suggest that the strategic value of air transportation infrastructure can be significant.

By adopting strategies that take advantage of the ability to react quickly to changes in the market, the value

of a piece of infrastructure can be increased. In this study, it was shown that a strategy of capacity delivery

based on small increments and short response times can yield more benefits than strategies that consider


large capacity increases and long response times; however, the ability to react quickly implies an a priori

investment. The difference between the expected value from the strategy with the ability to react quickly

and the bbusiness as usualQ strategy is the maximum that a rational investor should be willing to invest a

priori. The strategic value of reacting quickly increases if there is a moderate cost reduction in the delivery

costs as compared to strategies with long time to increase capacity.

A critical decision in any infrastructure capacity investments is when to begin work to add the capacity.

This study shows that there may be an optimal congestion threshold to begin adding capacity depending on

what strategy is being followed. In the specific airport example considered here, it was found that a

Congestion threshold of 75% should be the trigger for capacity enlargements if strategies based on small

capacity increments and 1 or 5 years to increase capacity are considered. The lesson for decision-makers is

that congestion delaysmust be addressedwith foresight. It may be too late to do something once congestion

becomes critical. Thus, depending on the expectations for future growth in air travel, the value of being able

to react quickly to changes in the market may be substantial.

The methodology described here can be expanded to evaluate the other two components of the strategic

value of air transportation infrastructure identified above, i.e. the ability to create and to stimulate markets.

The key is to introduce these processes into the system dynamics model, because once this is

accomplished, theMonte Carlo simulation and the evaluation of strategies as described above can be used.

The ability to create markets can be investigated by simulating the demand for air travel within a

competitive market that allows airlines to enter and leave the market served by the airport in the system

dynamics model. The entry/exit of airlines would depend, among other factors, on the congestion levels at

the airport and, thus, on the infrastructure investments. Moreover, the ability to stimulate markets can be

analyzed by allowing airlines in the competitive market model to expand and contract flight schedules.

Again, congestion levels at the airport would be a key element in the airline’s decision to alter its

operations.

References

[1] ICAO, Impact of Civil Aviation on State’s Economies. World-wide CNS/ATM Systems Implementation Conference, Rio

de Janeiro, 1998.

[2] Christine Kessides, A review of Infrastructure’s Impact on Economic Development, Infrastructure and the Complexity of

Economic Development, Springer Verlag, Berlin, 1996.

[3] Anil Kapur, Airport Infrastructure: The Emerging Role of the Private Sector, World Bank Technical Paper, vol. 313, The

World Bank, Washington D. C., USA, 1995.

[4] Richard Larson, Amedeo R. Odoni, Urban Operations Research, Prentice-Hall, New Jersey, 1981.

[5] Richard de Neufville, Amedeo Odoni, Airport Systems — Planning, Design and Management, McGraw-Hill, 2003.

[6] Steve Skinner, Alex Dichter, Paul Langley, Hendrik Sabert, Managing Growth and Profitability across Peaks and Troughs

of the Airline Industry Cycle — An Industry Dynamics Approach, in: Gail F. Butler, Martin R. Keller (Eds.), Handbook of

Airline Finance, MacGraw-Hill, 1999.

[7] Avinash K. Dixit, Robert S. Pindyck, Investment under Uncertainty, Princeton University Press, Princeton, New Jersey,

1994.

[8] USDOT, Form 41, US Department of Transportation, Bureau of Transportation Statistics, Office of Airline Information,

1979–2001.

[9] Peter Belobaba, Characteristics of Air Transportation Markets and Demand for Air Travel. Lecture notes from The Airline

Industry course taught at the Massachusetts Institute of Technology, 2001.

[10] FAA, Aviation Capacity Enhancement Plan, Federal Aviation Administration, Office of System Capacity, Washington,

D. C, 1992–2001, Available at: http://www1.faa.gov/ats/asc/ACE.html.

http://www1.faa.gov/ats/asc/ACE.html


Bruno Miller finished a doctoral degree in Air Transportation Systems at the Department of Aeronautics and Astronautics at

the Massachusetts Institute of Technology in Cambridge, MA, United States in June, 2005. He is now with Northwest Airlines

in Eagan, Minnesota, United States.

John-Paul Clarke is an Associate Professor in the School of Aerospace Engineering at the Georgia Institute of Technology

(Georgia Tech) where his research and teaching address issues of optimization and robustness in Aircraft and Airline

Operations, Air Traffic Management and the Environmental Impact of Aviation.

The hidden value of air transportation infrastructure

Documents

Transcript of The hidden value of air transportation infrastructure